Latest Contributions by EricKrause

It could be a couple of things.  First, regression kriging is a fairly complicated method for only 40 data points.  You may just not have enough data to apply such a complicated model.  Second, the residuals of your OLS model don't appear to display spatial autocorrelation.  Looking at the crossvalidation statistics and the empirical semivariances (the blue crosses in the semivariogram screen), your residuals seem to be independent and normally distributed.  Sampling more data may reveal more spatial patterns that can be used in kriging, but I don't see anything in those diagnostics that indicates the residuals of your OLS model need to be kriged at all. ... View more

If you have ArcGIS 10.1 and a Geostatistical Analyst license, you should consider using areal interpolation.  It works directly with polygons; there is no need to try to represent them as points. You can see a video that takes you through a typical areal interpolation workflow here. ... View more

I'm sorry, I messed up.  If you only change the auto flags for the major and minor ranges, it will calculate them assuming the other parameters are correct.  Since the other parameters were calculated for the very first dataset, they will not match new datasets, and the calculation of the major/minor range will get corrupted.  You'll need tell the model to recalculate everything before querying the ranges. I've attached an .zip file that contains an XML file that has all the flags changed in order to properly recalculate the anisotropy parameters.  Use this file as the model source in Create Geostatistical Layer.  Use the same XML path codes as above to query the major/minor ranges with Get Model Parameter. Also, if you want to query the angle of the major range, use this XML path code in Get Model Parameter: /model[@name='Kriging']/model[@name='Variogram']/model[@name='VariogramModel']/value[@name='Direction'] Sorry for the mess up, but this workflow should work now. ... View more

You need to do something similar to the steps in this topic. After you create a geostatistical layer using kriging with anisotropy turned on, you need to call the Set Model Parameter gp tool with your geostatistical layer as the model source.  You need to set the "auto" flag for the major and minor ranges to "true".  The code you'll need is this: Parameter XML path: /model[@name='Kriging']/model[@name='Variogram']/model[@name='VariogramModel']/value[@name='Range']/@auto; /model[@name='Kriging']/model[@name='Variogram']/model[@name='VariogramModel']/value[@name = 'MinorRange']/@auto Parameter value: true; true Save the output xml file in a convenient location. Now you need to use the Create Geostatistical Layer gp tool.  Use the xml file you saved above as the model source, and give it a new dataset.  Run the tool, and you'll get a new geostatistical layer where the major and minor ranges have been recalculated. You can then query the major and minor ranges by running Get Model Parameter gp tool twice.  For the model source, give it the geostatistical layer that you created from Create Geostatistical Layer, and use the following XML path code: /model[@name='Kriging']/model[@name='Variogram']/model[@name='VariogramModel']/value[@name='Range'] and /model[@name='Kriging']/model[@name='Variogram']/model[@name='VariogramModel']/value[@name = 'MinorRange'] The first bit of code will return the major range.  The second will return the minor range.  You only need to do Set Model Parameter once, and you can iterate through your datasets and keep re-using that xml file in Create Geostatistical Layer.  For each new geostatistical layer, use Get Model Parameter to query the major and minor ranges. I hope that was clear.  Let me know if you run into problems. ... View more

With no nugget, kriging is an exact interpolator.  If you use GA Layer to Points and predict back to the input point locations, you should get perfect predictions.  Specifically, the input point will get a weight of 1 and all other neighbors will get a weight of 0. In crossvalidation, you are throwing out the input point before predicting back to that input point location, so it can't just assign it a weight of 1 (because it has been thrown out of the dataset).  That is why an exact interpolator will still have crossvalidation errors. ... View more

Yes, you would want to compare the RMSE between EBK and IDW.  A large root-mean-square-standardized usually indicates the model is unstable.  The most common reason for this is because the Gaussian semivariogram is very unstable if the nugget is very small, compared to the sill.  Note that Stable with parameter=2 and K-Bessel with parameter=10 both correspond to the Gaussian semivariogram (it's a special case of both). EDIT:  Oh, I understand what you were asking.  It doesn't make much sense to compare RMS and average standard error from different models, but it is useful to compare them within the same model because if the difference between them is large, it indicates that the model may have problems. ... View more

I think the tool you're looking for is Central Feature.  It finds the feature with the smallest cumulative distance to all the other features. ... View more

What happens when you try to export the geostatistical layer to a raster? ... View more

Yes, you can change the extent of a geostatistical layer.  Right-click the layer in ArcMap's Table of Contents and choose "Properties."  Go to the Extent tab and specify the new extent. ... View more

Glad I could help.  Feel free to ask more questions if anything else comes up. ... View more

I've looked into why the Voronoi map is ignoring the top-right and lower-left cells.  The problem is that when you have gridded points, the Delaunay triangulation is not unique.  Since we define polygon neighbors at the triangulation step (the first step of creating the Voronoi polygons), our algorithm drops the top-right and lower-left neighbors.  A different but analogous implementation would drop the upper-left and lower-right polygons.  We could fix this by defining neighbors after the polygons are created, but this would slow down the tool.  We'll have to think whether this hit in performance is worth it, especially considering that Focal Statistics is specifically built to deal with gridded data. ... View more

I'll look into why the top-right and lower-left cells are being ignored, but after asking around, I think the tool you want to use is Focal Statistics.  It gives lots of options for defining cell neighbors, and you can calculate the standard deviation of these neighbors.  And I'm not surprised at the fairly weak R^2 between standard deviation and entropy.  Because entropy works with classified values (rather than raw values), the entropy map tends to be smoother.  Entropy also has a maximum, but standard deviation has no maximum.  So, two cells can have the same entropy value but still have very different standard deviations.  If you look at your scatterplot, you can even see this; there are clear vertical columns that all have the same entropy, but the standard deviations vary a lot.  This variance is what is pulling down the R^2. ... View more

The steps you've outlined are correct.  Unfortunately, there's no way to make a continuous OLS surface.  You'll need to make separate rasters for each resolution you want to test, but you only need to krig on the residuals once.  You can export the interpolated residual surface to any cell size and extent that you want. I now understand what you're trying to do with the correlation coefficient.  However, I don't think a correlation coefficient will work here because if you want to correlate a single point to the mean of its neighbors, you'll only be able to calculate a single coefficient for the entire surface (since you need repeated samples), so it won't help you in deciding which particular locations should be given preference.  The first thing that comes to mind is the Voronoi Map tool.  It's an interactive graphical tool, and if you use Standard Deviation, Entropy, or Interquartile Range, you'll get an estimate of the local variability.  A small local variability indicates that the predictions are more constant in that area, so they might be good candidates for new sites because the area can be better represented by a single value.  Note that you'll need to convert your rasters to points to run the tool. ... View more

Also, we heavily suggest using a projected coordinate system rather than a Lat-Long GCS.  Distance calculations get badly distorted when using Lat-Long, and this distortion will get propagated through all the kriging calculations. ... View more